2.1 NMR sample

The recombinant NC₅₅ protein was overexpressed from a recombinant plasmid (kind gift of M.F. Summers) and purified as previously described [11,12].

The RNA oligonucleotide mimicking the tRNA₃^Lys D-arm (24 nucleotides length) was purchased from Dharmacon Research with 2′-o-bis (acetoxyethoxy)-methyl (ACE) protection. The sample was deprotected by following the manufacturer's recommendations and dialyzed several times against deionised water. The pH was adjusted between 6 and 6.5 before freeze-drying.

The ¹⁵N-labelled NC₅₅ and the unlabelled RNA D-arm were prepared at about 2 mM and 10 mM, respectively, in an NMR buffer (25 mM Na-d₃-acetate pH 6.5, 25 mM NaCl, 0.1 mM BME, 0.1 mM ZnCl₂, 10% ²H₂O). The titration experiment is achieved by adding increased amount of RNA D-arm in a 0.5 mM sample of ¹⁵N-labelled NC₅₅. After RNA addition, the sample is immediately vortexed and kept at 30 °C during 30 min before NMR data acquisition.

2.2 NMR experiments

NMR experiments were carried out on a Bruker Avance DRX spectrometer operating at 600 MHz and equipped with a 5-mm triple-resonance gradient probe with z-axis gradients. All sequences used sine-bell shaped pulsed field gradients (PFG) along the z-axis. Acquisitions of data were performed at 303 K. All experiments were recorded with 512 points in the ¹H dimension with a spectral width of 4.6 ppm and 256 points in the ¹⁵N dimension with a spectral width of 30.0 ppm. NMR data were processed with NMRPipe/NMRDraw software [13]. A pure squared sine bell and a pure squared cosine bell were applied along t₂ and t₁, respectively. The data were zero-filled to 1024 points along t₁ and to 4096 along t₂ prior to Fourier transform. Spectra were analyzed with the software SPARKY (T.D. Goddard and D.G. Kneller, SPARKY 3, University of California, San Francisco).

The SE-HSQC experiment is a standard sensitivity-enhanced 2D ¹H/¹⁵N correlation via double INEPT transfer [14]. Frequency discrimination with retention of absorption mode lineshape is achieved by the echo anti-echo method [15]. Sensitivity enhancement is optimized for amide cross peaks by setting both reverse transfer periods to (1/2J). Heteronuclear decoupling during acquisition was achieved by GARP sequence.

The SE-CPMG-HSQC experiment (Fig. 1) is adapted from the SE-HSQC experiment in which the INEPT and reverse-INEPT periods are modified by the inclusion of a phase-modulated pulse train. This experiment is based on previously described HSQC experiments for the observation of exchange broadened signals [6]. The sequence was modified to optimize sensitivity enhancement for amide cross peaks by setting both reverse transfer periods to (1/2J). See caption of Fig. 1 for more details on NMR pulse sequence.

3.1 Biological context

The HIV-1 virus uses a cellular tRNA, the human tRNA₃^Lys as a primer to initiate the reverse transcription of the viral genome. In the virus, the 3′ end of the primer tRNA is strongly bound by 18 Watson–Crick base pairs to a complementary sequence of the viral RNA called the primer binding site (PBS). The first event in viral replication is thus the annealing of tRNA₃^Lys to the PBS, this step being promoted by the viral NC. Most of the known functions of the HIV-1 NC involve interactions with nucleic acids (for recent reviews, see Refs. [7–9]). The structure of the free NC consists in two zinc knuckle domains, connected by a flexible linker attached to flexible N- and C-terminal tails [12,16,17]. By NMR, it was shown that NC has an inherent conformational flexibility that could be crucial to allow its binding to different nucleic acid sequences and structures along the HIV-1 replication cycle [12,18]. In the currently available structures [11,19,20] of NC with nucleic acids, most of the specific contacts between the protein and the nucleic acids occur via the zinc knuckles, notably residues F₁₆ and W₃₇, interacting with the guanine-rich loop of RNA hairpin. We have shown by NMR of ¹⁵N-labelled tRNA₃^Lys that NC interacts preferentially with the acceptor and D stems and with the nucleotides involved in tertiary interactions between the D and T loops [21]. We also performed the reciprocal experiment, i.e. the NMR footprint of tRNA₃^Lys binding on ¹⁵N-labelled NC (data not shown). The signals are so broadened that roughly only the C-teminal amide group and the amino groups of side chains remain observable. Indeed, the multiplicity of NC binding sites on tRNA₃^Lys rendered analyses using the full-length tRNA intractable, even when stringency was raised by increasing salt concentration (data not shown). Reducing the target RNA length could decrease the number of non-specific sites and still have a biological interest. Indeed, studying separate hairpins cannot obviously mimic the entire tRNA, in particular at the level of the TΨC/D loops interaction. However, in the annealing process, there is a step where the TΨC/D loops interaction are broken and where the NC protein plays a crucial role [22]. Therefore, we have previously demonstrated that NC makes a specific interaction with the D hairpin mimicking the D-arm of tRNA₃^Lys (dissociation constant K_d = 28 nM at 100 mM NaCl). This interaction could be a key factor driving the unfolding process, thereby allowing complete annealing of tRNA₃^Lys to the PBS [10]. Fig. 2a shows the HSQC experiment of ¹⁵N-labelled NC with the tRNA₃^Lys D hairpin. The NMR signals are quite broad for the size of the complex (14.2 kDa) and may indicate that the NC in the context of the complex is involved in a chemical exchange process in the microsecond to millisecond time scale. This chemical exchange is attributable to the equilibrium between the free and bound forms of the protein.

3.2 Analysis of the exchange process and NMR optimization of HSQC spectra

In the presence of such chemical exchange, multidimensional NMR experiments suffer from dephasing of spin coherence during periods of evolution and coherence transfer by scalar coupling [2,3]. The result of chemical exchange during periods of evolution is a broadening of the resonances, while it leads to loss of magnetization during periods of coherence transfer and therewith of signal intensity. Therefore, resonances are broad, very weak or even unobservable (Fig. 2a). The loss of magnetization during the period of coherence transfer can be minimized by using a modified INEPT block, i.e. the CPMG-INEPT block first described by Mueller and co-workers [5]. The intensity of the signals involved in the chemical exchange process is increased by refocusing them during the coherence transfer period. We then compared CPMG-HSQC [6] and SE-CPMG-HSQC described in Fig. 1, since the gain of intensity from the sensitivity-enhanced HSQC could be counterbalanced by the increase of the pulse sequence length. As a result, the SE-CPMG-HSQC spectrum shows more intense signals than those from the CPMG-HSQC (data not shown).

The comparison of the performance of the SE-HSQC and SE-CPMG-HSQC experiment is given in Fig. 2. The CPMG-derived sequence shows improved sensitivity for almost all resonances, resulting in an average enhancement of about 40%. This result confirms that the exchange process lies in the microsecond to millisecond time scale. Moreover, some unobservable signals in the classic SE-HSQC experiment grow up in the SE-CPMG-HSQC and become observable (see signals 1, 2, 3 in Fig. 2). This could be of significant importance, as these signals, which show considerably higher intensity, belong to residues directly involved in RNA binding.

Fig. 3 shows the titration of ¹⁵N-labelled NC by increased amounts of D hairpin until reaching the equivalence. Two kinds of behaviour are observed: (1) a progressive variation of chemical shifts as the amount of RNA increases (see for instance the NMR signals of C₂₈ and C₄₉ – Fig. 3); (2) a disappearance at the substoichiometric level of NC signals and their reappearance close to the 1:1 ratio of NC:RNA (Fig. 3). These results are consistent with an exchange relaxation constant (R_ex) behaviour in the fast exchange regime. Indeed, for a simple model of two-site exchange process in the fast exchange limit, the exchange relaxation constant is given by Rex=papb(Δω2/kex), where k_ex is the exchange rate constant, Δω is the chemical shift difference between the two sites, and p_a and p_b the relative population of each site. For a complex with a dissociation constant more than 3 orders of magnitude smaller than the species concentrations, we can assume that the concentration of the complex in bound form is rather equal to the partner added in substoichiometric molarity. Thus, during the titration, for a fraction p of added RNA in the protein (p ∈ [0,1]), the fraction of the bound protein is p and the fraction of the free one is (1 − p) and so R_ex becomes Rex=p(1−p)(Δω2/kex). Even if these equations apply only to the fast exchange limit (k_ex > Δω), they provide useful qualitative information for the interpretation of the relaxation behaviour of a complex in a fast exchange regime. For instance, the R_ex maximum during the titration occurs for p = 0.5, and depends quadratically on the chemical shift difference between the free and the bound forms. Therefore, the NMR signals that experience the largest variation of chemical shifts between the free and bound forms of NC are also those that will experience the largest broadening during the titration due to the exchange regime.

We can reasonably think that these signals which present the largest differences in their chemical shifts between the free and bound forms of NC are those involved in the interaction with the RNA partner. Actually, they belong to amino acids surrounding the major aromatic residue of each zinc knuckle, i.e. F₁₆ and W₃₇, and to residues with hydrophobic side chain, i.e. I₂₄, A₂₅ and M₄₆ (Fig. 3). These residues have been identified to largely participate in the recognition by the NC of RNA encapsidation signals SL2 and SL3 [11,20]. Indeed, in these structures, each zinc knuckle forms a hydrophobic cleft by means of the hydrophobic side chains listed above and binds to an exposed guanine nucleobase of the RNA loop. Thus the NMR footprint of the RNA D-arm on the NC (Fig. 3) suggests that the recognition mode between the NC and the tRNA₃^Lys D-arm is similar to that of the encapsidation signals.